We present a novel self-correcting, high-speed optoelectronic probabilistic computer architecture that leverages source-device independent (SDI) quantum photonic p-bits integrated with robust electronic control. Our approach combines the intrinsic randomness and high bandwidth of quantum photonics with the programmability and scal- ability of classical electronics, enabling efficient and flexible probabilistic computation. We detail the design and implementation of a prototype system based on photonic integrated circuits and FPGA-based control, capable of implementing and manipulating 64000 logical p-bits. Experimental results demonstrate that our architecture achieves a flip rate of 2.7 x 10^9 flips/s with an energy consumption of 4.9 nJ/flip, representing nearly three orders of magnitude improvement in speed and energy efficiency compared to state-of-the-art magnetic tunnel junc- tion (MTJ) based systems. Furthermore, the SDI protocol enables real-time self-certification and error correction, ensuring reliable operation across a wide range of conditions and solving the problem of hardware variability as the number of p-bits scale. Our results establish quantum photonic p-bits as a promising platform for scalable, high-performance probabilistic computing, with significant implications for combinatorial optimization, machine learning, and complex system modeling.
We present a novel self-correcting, high-speed optoelectronic probabilistic computer architecture that leverages source-device independent (SDI) quantum photonic p-bits integrated with robust electronic control. Our approach combines the intrinsic randomness and high bandwidth of quantum photonics with the programmability and scal- ability of classical electronics, enabling efficient and flexible probabilistic computation. We detail the design and implementation of a prototype system based on photonic integrated circuits and FPGA-based control, capable of implementing and manipulating 64000 logical p-bits. Experimental results demonstrate that our architecture achieves a flip rate of 2.7 x 10^9 flips/s with an energy consumption of 4.9 nJ/flip, representing nearly three orders of magnitude improvement in speed and energy efficiency compared to state-of-the-art magnetic tunnel junc- tion (MTJ) based systems. Furthermore, the SDI protocol enables real-time self-certification and error correction, ensuring reliable operation across a wide range of conditions and solving the problem of hardware variability as the number of p-bits scale. Our results establish quantum photonic p-bits as a promising platform for scalable, high-performance probabilistic computing, with significant implications for combinatorial optimization, machine learning, and complex system modeling.
Probabilistic machine learning utilizes controllable sources of randomness to encode uncertainty and enable statistical modeling. Harnessing the pure randomness of quantum vacuum noise, which stems from fluctuating electromagnetic fields, has shown promise for high speed and energy-efficient stochastic photonic elements. Nevertheless, photonic computing hardware which can control these stochastic elements to program probabilistic machine learning algorithms has been limited. Here, we implement a photonic probabilistic computer consisting of a controllable stochastic photonic element - a photonic probabilistic neuron (PPN). Our PPN is implemented in a bistable optical parametric oscillator (OPO) with vacuum-level injected bias fields. We then program a measurement-and-feedback loop for time-multiplexed PPNs with electronic processors (FPGA or GPU) to solve certain probabilistic machine learning tasks. We showcase probabilistic inference and image generation of MNIST-handwritten digits, which are representative examples of discriminative and generative models. In both implementations, quantum vacuum noise is used as a random seed to encode classification uncertainty or probabilistic generation of samples. In addition, we propose a path towards an all-optical probabilistic computing platform, with an estimated sampling rate of ~1 Gbps and energy consumption of ~5 fJ/MAC. Our work paves the way for scalable, ultrafast, and energy-efficient probabilistic machine learning hardware.
This paper proposes a novel approach to enhancing the reliability of digital circuits by leveraging quantum error correction (QEC) techniques. Current digital circuits are susceptible to bit flips and other errors, leading to system failures. The core idea is to embed QEC codes directly within the circuit architecture, allowing for the real-time detection and correction of these errors. This represents a significant departure from traditional fault-tolerance strategies, which typically rely on external redundancy. We outline the theoretical framework, detailing the implementation of specific QEC codes and their integration into a digital circuit. The potential impact of this approach – inherently self-correcting circuits – is explored, arguing for a dramatic improvement in system robustness and longevity. The primary focus is on the conceptual design and feasibility of this method, without requiring experimental validation. The key contribution lies in directly applying the principles of QEC, originally developed for quantum computing, to the design of classical digital systems.
The rapid rise of quantum computing and quantum internet highlights the irreplaceable importance of high-speed, scalable quantum randomness certification (QRC). However, existing QRC implementations are constrained by quantum measurement limits and integration bottlenecks, leaving both high-speed operation and monolithic implementation unrealized. Here, we demonstrate a monolithically integrated photonic–electronic on-chip source device-independent (SDI) quantum random number generator (QRNG) that achieves a secure bit rate of 583.2 Gbps within a compact 3 mm ×11 mm ×11 mm footprint, marking an advance in both integration and speed, nearly 30 times faster than the best existing integrated SDI implementations. By harnessing measurement-induced quantum uncertainty as a resource for quantum certification, this study resolves the challenge of uniting high-speed certified entropy bounds with full monolithic integration in SDI QRC. A universal entropy bound incorporating high-bandwidth physical constraints enables accurate entropy estimation under integrated conditions. This theoretical framework supports a four-path multi-interference measurement scheme that improves entropy yield and fills a critical gap in digital-layer security under coherent attacks. These contributions are experimentally embodied in a compact photonic–electronic chip that performs NIST-compliant secure entropy extraction, demonstrating the feasibility of combining high-speed performance with full system integration. Compatible with standard optical communication infrastructure, this study sets a new benchmark for scalable high-speed QRC, enabling robust security and broad applicability across quantum computing, the internet, and cryptography.
We present photonic quantum computing architectures that can deal with both probabilistic (heralded) generation of single photons and probabilistic gates without making use of coherent switching. The only required dynamical element is the controllable absorption of all photons in a given mode. While the architectures in theory scale polynomially in the resources required for universal quantum computation, as presented their overhead is large and they are illustrative extreme points in the configuration space of photonic approaches, rather than a recipe that anybody should seriously pursue. They do, however, prove that many things presumed necessary for photonic quantum computing, in fact are not. Of potentially independent interest may be that the architectures make use of qubits which have many possible microstates corresponding to a single effective qubit state, and the technique for dealing with probabilistic operations is to, when necessary, just enlarge the set of such microstates to incorporate all possibilities, while making heavy use of the subsequent ability to `coherently erase' which particular microstate a given qubit is in.
Qyn is a project that presents the design and simulation of an FPGA-based three qubit bit-flip quantum error correction (QEC) architecture that models the complete error correction control loop used in quantum computing systems. The proposed design of this project includes logical state encoding, probabilistic single qubit error injection using a pseudo-random noise source, syndrome extraction, deterministic error correction, and logical decoding via majority voting. Extensive Monte Carlo-style simulation demonstrates a clear reduction in logical error rate relative to the physical error rate, validating the effectiveness of the implemented correction logic under stochastic noise conditions. In addition, a quantum machine learning (QML) block was implemented using fixed-point arithmetic to illustrate how low-latency classical inference can be combined with error correction hardware. Although operating in a fully digital domain and limited to bit flip errors, Qyn demonstrates the feasibility and scalability of low-power FPGA-based control architectures for quantum error correction and hybrid quantumclassical systems.
Quantum Error Correction (QEC) is often idealized as hardware-agnostic, yet its practical implementations must be tailored to the underlying system. The nuclear and electron spins of donor atoms in silicon constitute a physical platform that offers long coherence times, a well-characterized Hamiltonian, precise quantum control, and a readout mechanism that uses the electron as an ancilla. In this thesis, we reconsider the fundamentals of QEC, from gates to architecture, showing how they can be tailored to donor spins, while exploring different architectural paradigms using donor spins. Motivated by the hardware’s native Hamiltonian and ancilla-assisted readout, we design Single-Step Parity Check (SSPC) gates that perform parity checks directly, bypassing 1- and 2-qubit decompositions and improving circuit error tolerance. We developed a numerical workflow using experimental process matrices to evaluate these operations and quantified cumulative losses in a two-body parity-check circuit by tracking 1- and 2-qubit gate error rates on a two-phosphorus, one-electron silicon chip. GRAPE-optimized pulses were generated for both standard and SSPC-based circuits on the same platform, showing SSPC gates outperform and offer hardware-aware QEC implementations. We next explore antimony, a high-spin donor, as a candidate for photonic quantum computing. With a nuclear spin of 7/2, antimony has eight energy levels which, when coupled to a microwave cavity, can emit a single-photon across eight modes. This supports third quantization, where (near-)deterministic photon entanglement is achieved by spreading single photons across many modes and sending each mode to a party, creating entanglement without photon–photon interactions or probabilistic gates. We proposed an experiment for the simplest realization of third quantization, where an antimony donor produces two independent 8-mode photons, distributed to parties, creates a Bell state between 56 random pairs with 87.5% efficiency. We then explored high-dimensional fusion operations for fusing qudit cluster states. Focusing on Type-II fusion, a central component of measurement-based quantum computing, we derived the most efficient high-dimensional protocol known to date — a 723-fold improvement over the best previously known protocol in the case of d = 5. To further enhance success probability of fusion, we introduced extra-dimensional corrections framework. We unified these insights to design two approaches for qudit graph state generation using antimony donors. One is for the direct generation of arbitrary graph states from coupled antimony donors, where two antimony atoms share a single electron, forming 128-dimensional Hilbert space, while the other focuses on the generation of linear graph states from a single antimony donor, followed by high-dimensional fusion operations to construct graph states in arbitrary shapes. We compared the two schemes in terms of their capabilities and potential for scalability.